Play2Vec web-based service for similar play retrieval

• Main Author: John, Benedict
• Other Authors: -
• Format: Final Year Project
• Language: English
• Published: Nanyang Technological University 2019
• Subjects: Engineering::Computer science and engineering::Computing methodologies; Engineering::Computer science and engineering::Computing methodologies::Artificial intelligence
• Online Access: https://hdl.handle.net/10356/136552
• Institution: Nanyang Technological University

With the advent of technology in sports such as advanced 3D positioning tracking system and datasets, there is an unprecedented of such high-definition information for sports evaluation, coaching, talent management and study of sports psychology. However, the ability to interpret such data in a coherent and useful manner as part of sports analysis is an area for development. To achieve this, Play2vec was developed to allow deep learning approach to learn the representations of sports play. This report highlights how the Play2vec approach is used to provide a web-based service which provides data analytics to render soccer analysts to compare and analyse between similar play types which can subsequently be used by coaches. For this project, an experimental and interim web-based service was developed to demonstrate a slice of the Play2vec deep learning representation model to retrieve similar sports play focused on soccer. The web-based service aims to provide sports analysts with technical experience a platform to evaluate game (soccer) play which can be distilled and used by coaches in performance management, talent management and evaluation of sports psychology in soccer play.